Software-defined networking (SDN) provides fine-grained network control and monitoring that simplifies network management. Unfortunately, upgrading existing enterprise networks, comprised of numerous "legacy" switches, to SDN is often cost-prohibitive. We argue that it is possible to achieve most of the benefits of a fully deployed SDN at a fraction of the cost by strategically replacing only few legacy switches with - or introducing a few - new SDN-capable switches in a legacy network, thus creating a hybrid network. We present Magneto, a unified network controller that exerts SDN-like, fine-grained path control over both OpenFlow and legacy switches in hybrid networks. Magneto i) introduces magnet MAC addresses and dynamically updates IPto-magnet MAC mappings at hosts via gratuitous ARP messages for visibility and routing control; and ii) uses the ability of SDN switches to send "custom" packets into the data plane to manipulate legacy switches into updating forwarding entries with magnet MAC addresses for enhanced routing flexibility. Our evaluation on a lab testbed and through extensive simulations on large enterprise network topologies show that Magneto is able to achieve full control over routing when only 20% of network switches are programmable, with negligible computation and latency overhead.
|Original language||English (US)|
|Title of host publication||SOSR 2017 - Proceedings of the 2017 Symposium on SDN Research|
|Publisher||Association for Computing Machinery, Inc|
|Number of pages||13|
|State||Published - Apr 3 2017|
|Event||2017 Symposium on SDN Research, SOSR 2017 - Santa Clara, United States|
Duration: Apr 3 2017 → Apr 4 2017
|Name||SOSR 2017 - Proceedings of the 2017 Symposium on SDN Research|
|Other||2017 Symposium on SDN Research, SOSR 2017|
|Period||4/3/17 → 4/4/17|
Bibliographical noteFunding Information:
This research was supported in part by NSF grants CNS-1411636, CNS 1618339 and CNS 1617729, DTRA grant HDTRA1-14-1-0040 and DoD ARO MURI Award W911NF-12-1-0385.
© 2017 ACM.
- Hybrid SDN
- MAC learning and forwarding